Distributed DAS Storage Orchestration for Non-Centralized Rebuilds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data storage systems face inefficiencies and high costs due to reliance on centralized storage arrays, which require extensive infrastructure, are prone to errors, and lack flexibility in managing distributed resources, leading to suboptimal performance and maintenance challenges.
Innovation Solution
A data storage system utilizing unexploited Direct Attached Storage (DAS) resources across multiple servers, orchestrated by a control plane (CP) to manage storage stacks, enabling efficient data management, redundancy, and flexibility without the need for dedicated storage arrays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If centralized storage arrays are used to provide data storage and backup capabilities, then data storage management is centralized and controlled, but infrastructure complexity and cost increase significantly
Solution Approach 1:
The patent segments the centralized storage array into distributed storage components across multiple servers. Each server manages its own storage locally (Direct Attached Storage), eliminating the need for a single centralized storage array and its associated complex infrastructure while maintaining centralized management capabilities through software orchestration.
Solution Approach 2:
The patent extracts storage functionality from the centralized storage array and distributes it to individual servers. By taking out the storage function from the centralized controller and embedding it in each server, the system reduces infrastructure complexity while maintaining management capabilities.
2Reliability
If traditional storage arrays provide redundant arrays of independent disks (RAID) for data protection, then data redundancy is achieved, but performance penalty occurs on write operations due to parity calculations
Solution Approach 1:
The patent replaces traditional mechanical RAID parity calculations with software-based error correction and redundancy mechanisms. Instead of using hardware RAID controllers that perform synchronous parity calculations (which bottleneck write performance), the system uses software error correction codes and distributed redundancy that can be applied asynchronously, improving write performance while maintaining data protection.
Solution Approach 2:
The patent changes the redundancy parameter from synchronous parity-based (traditional RAID 5/6) to asynchronous error correction-based. This parameter change allows write operations to complete without waiting for parity calculations, significantly improving write performance while maintaining equivalent or superior data protection capabilities.
3Reliability
If storage arrays use dedicated hardware for system and application operations, then storage services are provided, but flexibility in managing distributed resources is reduced
Solution Approach 1:
The patent makes each server universal by embedding storage functionality directly in every server node. Each server can simultaneously function as both a compute node and a storage node, providing multi-functionality. This universal design enables flexible management of distributed resources, as any server can be dynamically assigned storage roles based on workload demands, unlike dedicated storage arrays with fixed hardware configurations.
Solution Approach 2:
The patent introduces dynamic resource allocation where storage capacity and roles can be dynamically assigned to different servers based on real-time needs. The system can dynamically add or remove storage nodes, change storage configurations, and redistribute data without physical reconfiguration, providing adaptability that static dedicated hardware cannot achieve.
4Device complexity
If single-core processors are used for parity calculations in RAID arrays, then hardware simplicity is maintained, but performance is limited and cannot utilize multi-core processors
Solution Approach 1:
The patent replaces single-core parity calculation mechanisms with multi-core parallel processing capabilities. By substituting the traditional single-threaded RAID controller approach with software-based error correction that leverages multiple CPU cores, the system achieves significantly higher productivity while maintaining acceptable hardware simplicity.
Data Source
AI summary
A data storage method and system comprising: at least one target server that comprises a storage media and configured to run an operating system designated to host data accessible over a data plane (DP) network, at least one initiator server comprises a storage media and configured to run an operating system designated to access and expose a remote resource/s over the data plane (DP) network, and at least one orchestrator configured to interact with each of said servers and designated to control a control plane (CP) of said DP network, wherein the combined storage medias form a shared drive storage stack, wherein the orchestrator is configured to utilize the CP in order to monitor and inspect the operability and condition of the shared drive storage stack, and further configured to monitor, record and inspect the proper operation of block device/s stored within said shared drive storage stack, wherein the orchestrator is configured to identify a malfunction affecting the shared drive storage stack, and wherein the orchestrator is configured to conduct a non-centralized rebuild procedure by managing a simultaneous and coordinated utilization of at least one operatable remote storage media that form the shared drive storage stack.